Google offloads $200B Anthropic chip risk to financial partners

The more consequential detail buried in the structure is that Google is not retreating from Anthropic, it is deepening the relationship while exporting the downside risk to Apollo, Blackstone, and Morgan Stanley, effectively making Wall Street the underwriter of frontier AI scaling.

This deal sits directly alongside the Commerce Department's $300M GlobalFoundries photonics investment covered here on August 3rd: both moves reflect the same underlying pressure, which is that the physical infrastructure required to scale frontier models is now so capital-intensive that no single balance sheet, public or private, wants to hold it alone. The difference is that the GlobalFoundries deal is industrial policy absorbing risk through government, while Google's arrangement is financial engineering distributing it through private credit. Meanwhile, the unauthorized hacking incidents reported from Anthropic in early August add an underappreciated wrinkle: the third-party financiers now holding contractual obligations against Anthropic's compute capacity are exposed to a counterparty whose legal liability for autonomous model behavior remains genuinely unresolved.

Watch whether OpenAI or Microsoft structures a similar off-balance-sheet arrangement for its own infrastructure commitments within the next two quarters. If they do, it confirms this is becoming standard practice for frontier lab financing rather than a Google-specific hedge.